Alkyl Chain Length‐Dependent Amine‐Induced Crystallization for Efficient Interface Passivation of Perovskite Solar Cells

Efficient surface passivation of perovskite solar cells (PSC) using treatment with ammonium salts is demonstrated as an efficient method to enhance the device performance, owing to the affinity between the amine group and [PbI6]4− octahedron. However, due to their high solubility in polar solvents (DMF/DMSO), ammonium salts are more difficult to use in passivation of the interface between the electron transport layer and perovskite thin film in n‐i‐p structured PSCs. In this report, this work successfully links the amine group with a fullerene through a series of increasing carbon chain length, from two to twelve methylene units (FC‐X, X = 2, 6, 12), and then introduce the synthesized molecules as interface passivation layers into SnO2‐based planar n‐i‐p PSCs. Results show that the interface passivation effect is highly dependent on the side‐chain length, and the longer chain length amine‐functionalized fullerene is more beneficial for the device performance. A power conversion efficiency as high as 21.2% is achieved by using FC‐12. The surface energy, perovskite crystallite size and electron transfer capacity correlate with the linker chain length. This work develops an amine‐induced anchored crystallization of perovskite to unravel the mechanism of this passivation effect. As expected, enhanced device stability is also observed in the FC‐12 passivated PSCs.


Introduction
Photovoltaic devices based on organicinorganic lead halide perovskite light absorbing materials have drawn increasing worldwide attention owing to the rapid rise in device photon-electron conversion efficiency since 2012. [1][2][3] The certificated power conversion efficiency (PCE) has surpassed 25%, and closes to the value of 26.7% achieved for silicon photovoltaic devices. [4] As a consequence, hybrid perovskite solar cells (PSCs) have been considered as the next generation photovoltaic technology with the most potential.
However, there remain distinct inadequacies of PSCs. For example, there are at least five interfaces (transparent electrode/ electron transport layer (ETL), ETL/perovskite, perovskite grain boundary, perovskite/hole transport layer (HTL), HTL/ metal electrode) in a typical n-i-p PSC architecture. [5] Some losses are expected when the photo-generated carriers transport from one layer to another. Moreover, due to the low temperature synthesis process and perovskite crystallographic features, numerous defect sites, especially delocalized organic amine defects, exist on the perovskite thin film surface and grain boundaries. These defect sites act as recombination centers which limit the device performance. [6] To address the above issues, interface passivation technology has been developed in the last few years. Proper passivation reduces the defect states on the perovskite thin film surface and at grain boundaries and, in addition, the use of an orchestrated passivation agent is able to reduce the band energy mismatch between perovskite thin film and other functional layers, thus reducing the open-circuit voltage loss. [6,7] To date, the most efficient PSCs reported were achieved through a surface passivation strategy by post-processing the interface between the perovskite and hole transport layer with a long-chain ammonium salt. [8] It seems likely that the passivation strategy will still play a critical role for developing efficient and environmentstable perovskite solar cells in the future.
The reported passivation agents cover a range of materials, including polymers, organic small molecules, alkali Efficient surface passivation of perovskite solar cells (PSC) using treatment with ammonium salts is demonstrated as an efficient method to enhance the device performance, owing to the affinity between the amine group and [PbI 6 ] 4− octahedron. However, due to their high solubility in polar solvents (DMF/DMSO), ammonium salts are more difficult to use in passivation of the interface between the electron transport layer and perovskite thin film in n-i-p structured PSCs. In this report, this work successfully links the amine group with a fullerene through a series of increasing carbon chain length, from two to twelve methylene units (FC-X, X = 2, 6, 12), and then introduce the synthesized molecules as interface passivation layers into SnO 2 -based planar n-i-p PSCs. Results show that the interface passivation effect is highly dependent on the side-chain length, and the longer chain length aminefunctionalized fullerene is more beneficial for the device performance. A power conversion efficiency as high as 21.2% is achieved by using FC-12. The surface energy, perovskite crystallite size and electron transfer capacity correlate with the linker chain length. This work develops an amine-induced anchored crystallization of perovskite to unravel the mechanism of this passivation effect. As expected, enhanced device stability is also observed in the FC-12 passivated PSCs. metal salts. [6,7] Among them, two material categories that are commonly utilized for passivating the interface in PSCs are fullerene derivatives, [9][10][11][12][13][14][15] and ammonium base salts. [8,[16][17][18] The fullerene group is a benign hydrophobic electron acceptor, used to improve the electronic transport efficiency between perovskite thin film and the electron transfer layer, as well as enhance the humidity resistance. On the other hand, ammonium-based salts have affinity for coupling with the [PbX 6 ] 4− octahedron, with a subsequent reduction in the exposed Pb defects in the perovskite crystalline surface. [19,20] However, most interface passivation using ammonium salts is usually applied between the perovskite and hole transport layer, [8,17,21,22] as it is difficult to passivate the interface between the electron transport layer and perovskite thin film in n-i-p structured PSCs owing to the solubility of the ammonium salts or precursor amines. [23][24][25][26][27] In this report, we designed and synthesized a series of fullerene derivatives with amine terminated alkyl substituents, with two, six, or twelve methylene units (FC-2, FC-6, and FC-12). By applying the aminefunctionalized fullerenes as an interface passivation layer between the SnO 2 electron transport layer and the perovskite photo-active thin film in PSCs, we found that the passivation effects are related to the side-chain length, and not all the synthesized amino-terminal fullerene molecules could introduce positive passivation outcomes. Only the amine functionalized fullerene with the longer carbon chain (FC-12) demonstrated a significant passivation role with an improved photovoltaic performance.

Results and Discussion
The amine functionalized fullerene derivatives were synthesized by an amidation of PC 61 BM with the relevant C2, C6, or C12 diamine following reported procedures [28] and their structures are displayed in Figure 1a, with synthetic details presented in the Supplementary Information. Hereafter, the derivatives are abbreviated as FC-X, where X = 2, 6, or 12.
The perovskite solar cell device architecture used in this work is presented in Figure 1b. A typical planar n-i-p device using SnO 2 as the electron transport layer and the perovskite crystalline layer was prepared through a two-step subsequent deposition method. [29] For comparison, the FC-X solution was spin coated between SnO 2 and perovskite as a passivation interface, as demonstrated in the process flow diagram in Figure S1, Supporting Information.
The current density-voltage (J-V) plots under AM 1.5 irradiation are shown in Figure 1c and the corresponding photovoltaic parameters are summarized in Table 1. The best PCE presented by the device without passivation was 19.5% (17.1 ± 1.8% average based on more than 20 devices, the same as below), with an open-circuit voltage (V oc ) of 1.08 V, short-circuit current density (J sc ) of 23.78 mA cm −2 , fill factor (FF) of 75.8%. After FC-X molecule layer insertion, the device performance shows an effect of the carbon chain length, that is, the X value. The devices passivated by FC-12 molecules demonstrate a significant enhancement of efficiency, in which the champion device shows the PCE of 21.2% (average value 20.2 ± 0.6%), with a V oc of 1.11 V, J sc of 24.19 mA cm −2 and FF of 79.2%. Compared to the control device, all the J-V parameters in FC-12 devices are improved, with an average PCE increase of more than 10%.
However, using FC-2 or FC-6 as an interlayer showed only a mediocre performance, with no notable improvement over the control sample. Specifically, the FC-2-based devices present the lowest PCE of 18.8% (average 17.6 ± 1.2%), a V oc of 1.04 V, J sc of 24.00 mA cm −2 , and FF of 75.8%. The devices treated by FC-6 are similar to the FC-2 devices, among which the best PCE is 19.2% (average 17.3 ± 1.1%), with a V oc of 1.05 V, J sc of 24.04 mA cm −2 , and FF of 76.1%.
Statistics box profiles of each photovoltaic parameter and the distribution curves are plotted in Figure 1d-g. Clearly, the distribution of each parameter decreased with carbon linker chain www.advmatinterfaces.de extension. Devices passivated by FC-12 are highly reproducible, which might result from the improved uniformity of the perovskite absorber layer. [30,31] Although FC-12 devices exhibit the best values for all the parameters, it should be noted that the variation in the J sc and V oc are not consistent. The J sc increases after the FC-X interlayer insertion, while the V oc dropped in FC-2 and FC-6 devices, with a similar trend in the PCE values mentioned above. Next, we also studied the hysteresis behavior of the devices by plotting J-V curves of reverse (V OC to J SC ) and forward scan (J SC to V OC ) for the devices with different fullerene interlayers as shown in the supporting information Figure S4, Supporting Information. We used a hysteresis index (H-index) to study the hysteresis behavior of perovskite solar cells, and the device characterization parameter along with H-index is summarized in Table S1, Supporting Information. The perovskite device with pure SnO 2 layer shows higher hysteresis with an H-index of 11.5% With the fullerene passivation layer, the hysteresis of J-V curve is reduced with increasing carbon chain length of the fullerene interlayer. The perovskite device with FC12 interlayer shows very low hysteresis of 1.6%, which is attributed to the efficient charge transport at the interface due to the efficient passivation of SnO 2 layer as well as, reduced trap density and enhanced crystallinity of perovskite layer.
In general, for a solar cell device with a specific absorber material and architecture, elevation of J sc or V oc can be modified through separate strategies. Typically, improvement of J sc is mainly determined by the optimization of electron extraction efficiency. On the other hand, enhancement of V OC usually results from the significant suppression of nonradiative charge recombination, that is, a passivation effect, which will be discussed later.
Numerous previous studies have reported that fullerene derivatives could passivate either the ETL or perovskite crystalline grains in PSCs. [32][33][34][35] It can be concluded from the variation of V oc data that the length of the linking carbon chain in FC-X molecules plays a "switching function" on the interface passivation effect: The photovoltaic performance was inhibited for a short carbon chain (FC-2), and the negative effect is mainly caused by the drop of V oc . The PCEs are almost equivalent as the carbon chain length extends to 6 and then, high efficiency devices are successfully fabricated when FC-12 molecule is used as the passivation layer.
Several efforts have been carried out to characterize the perovskite thin film before and after FC-X treatment, for the purpose of exploring the underlying mechanisms of the passivation process. Comparative UV-Vis absorption spectra were collected to identify if there is any absorbance variation of the perovskite layer after FC-X molecule interaction. In Figure 2a, the absorbance edge near 790 nm is the characteristic absorption band of the di-cation perovskite used in this work. [36] All the spectra overlap with each other quite well and there is no absorbance change nor wavelength shift.
The electron extraction dynamics of the perovskite films were investigated to obtain more insights into how the mechanism of PSC device performance varies with each fullerene derivative treatment. Photoluminescence (PL) plots ( Figure 2b) present the fluorescence quenching of the perovskite thin film when it interacts with various electron transport layers. Compared to the pristine perovskite thin film, the PL intensity was quenched about 56% by SnO 2 . The FC-2 and FC-6 interlayer showed a similar quenching effect, in which the PL intensity dropped by ≈75%. Notably, more than a 90% PL intensity reduction was observed with the FC-12 sample, demonstrating the FC-12 has an appreciably higher electron extraction ability and associated enhancement of nonradiative decay rates.
These PL intensity variations suggest direct evidence that the FC-X treatments are improving the electron transport injection from perovskite into ETL in a manner consistent with the observed changes in device performance: J sc slightly improved along with the FC-X treatment, nevertheless, FC-2 interlayer results in a decrease in PL intensity yet a concomitant decrease in device performance. Meanwhile, the PL peak position has a red shift from 790 to 801 nm after FC-12 treatment (Figure 2c), which might arise from the self-absorption induced by grainsize increase. [37,38] A plausible mechanism is proposed to explain the performance variation induced by the FC-X interlayer. As displayed in Figure 3, the wettability of the two terminal groups on the FC-X is quite different: -NH 2 is a rather hydrophilic group with appreciable affinity with PbI 2 . [19,25] The -NH 2 acts as nucleation sites during the PbI 2 spin coating process. At the opposite chain end, the fullerene cage is highly hydrophobic. Therefore, a simplified PbI 2 Nucleation-Crystallization-Growth process could be described as follows: PbI 2 in the precursor solution anchors to the -NH 2 and crystal seeds are generated; [39,40] then the PbI 2 grain grows; finally, the crystalline grain extends until reaching the C 60 cage. Thus, the size of crystalline PbI 2 is roughly determined by the distance between the -NH 2 and C 60 groups, which is the carbon chain length. Typically, the perovskite crystallite size plays an important role in the PSC's photovoltaic performance, a larger crystal grain size is usually associated with fewer populated trap sites and facile carrier diffusion length, and these features are beneficial to the device performance. From the above hypothesis and discussion, it can be proposed that FC-12 with the longest -NH 2 to C 60 distance will be prone to form a larger PbI 2 /perovskite grain size and thus lead to a higher PCE of the fabricated devices. A bimolecular layer model was established to elaborate the anchor functions of amine terminated group on SnO 2 and the perovskite thin film. As displayed in Figure 3, during spin coating of FC-12 on SnO 2 thin film, a bilayer with opposite orientations is generated: a bottom FC-12 layer with amine toward SnO 2 and the top layer with fullerene toward the inside, leading to the amine groups being exposed to the perovskite thin film. The amine terminated groups with a Lewis base feature are able to interact with un-coordinated Sn, which favors the transfer of photoexcited electrons from the perovskite film to the SnO 2 ETL. [11] On the other side, lone electron pairs on the amine effectively "bond" with the perovskite layer by exploiting the associated intermolecular interaction between nitrogen and un-coordinated lead divalent cations. [8] Therefore, from this bilayer molecule arrangement, the surface defects on both SnO 2 and perovskite crystal grains could be passivated to some degree by the anchor functions of FC-12.
An issue that remains is that the FC-2 and FC-6 devices have quite similar charge extraction and photovoltaic properties, although the carbon number in the linking chain of FC-6 is triple that of FC-2. Molecular geometries could explain this apparent anomaly. Due to interatomic interactions, the spatial configuration of the carbon chain is not linear. Spatial bending and folding contribute to minimizing the molecule energy. The Gaussian program was used to optimize the molecular geometries of FC-X molecules. The results demonstrate the distinct folding of the -(CH 2 ) 6 -chain with calculated distances between NH 2 and C 60 of 0.81, 1.46, and 2.13 nm for FC-2, FC-6, and FC-12 molecule, respectively. The distance values for FC-2 and FC-6 are quite close and similar to that of the C 61 cage (0.71 nm).
To verify the validity of the above hypothesis, contact angle measurements were performed on ITO/SnO 2 substrates with and without FC-X interlayers, with the obtained images displayed in Figure S3, Supporting Information. Contact angles show a decreasing trend with the carbon chain extension. The FC-2 surface is most hydrophobic, with the contact angle of 77.5°, the FC-6 (73.5°) is slightly lower but close to that of FC-2, while FC-12 presented the highest hydrophilicity among the three molecules, with a contact angle of 57.5°, indicating that more amine groups are exposed, consistent with the bilayer model. The change of contact angles along with carbon chain length agrees well with the plausible mechanism discussed above. In addition, we investigate the interaction between SnO 2 and fullerene derivatives, using X-ray photoelectron spectroscopy (XPS). The high-resolution XPS spectra of Sn3d 3/2 and Sn3d5/2, and N1S peak were obtained for pure SnO 2 and SnO 2 /fullerene derivative thin films and the results are given in supporting information Figure S5, Supporting Information. As shown in Figure S5a, Supporting Information, strong peaks at 486.3 and 494.7 eV attributed to Sn 4+ ions indicate the formation of SnO 2 . Comparing to the pure SnO 2 layer, amine-functionalized fullerene passivated SnO 2 film exhibit an Sn3d peak shift toward higher binding energy suggesting that the chemical change of the SnO 2 film which can be attributed to the amine bonding to SnO 2 . In addition, Figure S5b, Supporting Information exhibits a distinct change of N1s intensity at 400 eV, which is attributed to NSn or NC bonds. The SnO 2 /FC12 www.advmatinterfaces.de film shows a higher intensity of N1s peak, suggesting that both SnN and CN bonds exist due to the favorable orientation of the fullerene layer to facilitate the amine-anchored crystallization, as shown in Figure 3.
We also characterized the surface morphology root-meansquare roughness (R ms ) by atomic force microscopy (AFM), with the surface morphology shown in Figure S2, Supporting Information. A gradient decrease was observed after FC-X coating, with the SnO 2 thin film coated by FC-12 interlayer showing the smallest R ms (0.88 nm, compared to 1.42 nm in that of the SnO 2 thin film), which is essential for efficient electron transfer and suppressed recombination. [41] In addition, the contact angle is a measure of surface energy, which impacts the grain size during the growth of the PbI 2 microstructure and further the perovskite crystalline grains. [42] The prediction about the PbI 2 and perovskite grain size alternation along with various interlayers is evident in the SEM measurements in Figure 4. The performance of PSCs is highly dependent on the quality of the perovskite film, including surface coverage, grain size, and crystallinity. For the SnO 2 substrates without FC-X treatment, the generated PbI 2 thin film consists of a porous surface with irregular-shaped grains of about several tens to hundreds of nanometers in size (Figure 4a). The PbI 2 thin film morphology after FC-2 treatment kept the similar mesoporous structure as that of bare SnO 2 . Yet due to its high surface energy, PbI 2 grains with increased pore density were observed (Figure 4b). The presence of a pore structure induces poor surface coverage of PbI 2 on the substrate, leading to pinholes existing on the perovskite thin film surface, which is detrimental to device performance. Following FC-6 treatment, larger PbI 2 crystals close to 1 micro meter appear with a significant reduction in porosity (Figure 4c). In addition, some hexagonal sheet grains with regular shape form, indicative of highly crystalline material. Notably, the PbI 2 film following treatment with FC-12 is quite dense, there are no visible pores nor macroscopic defects to be found in the SEM image (Figure 4d). Most of the crystalline grains have a textured feature (Figure 4h), for which morphology the crystallinity is improved and the diffuse reflectance losses reduced. [43] Furthermore, the morphology of the obtained perovskite thin films present similar variation tendency as the PbI 2 thin films discussed above. More than that, it should be highlighted that the FC-X treatment could inhibit the generation of decomposed PbI 2 grains during heat-annealing of the perovskite in air. Crystal grains with bright contrast are generally considered to be PbI 2 . [29] By comparing the amounts of the bright grains in Figure 4e-h, decomposed PbI 2 is reduced considerably after FC-X layer insertion. It has been demonstrated that a certain amount of excess PbI 2 is favorable for the charge separation and stability of perovskite thin films, while superabundant PbI 2 residue is detrimental for the device. [44,45] The advantage might result from the anchor effect between -NH 2 group and PbI 2 induced grain dimension growth. [8] The crystallinity of the perovskite thin films was also characterized by the Grazing-Incidence Wide-Angle X-ray Scattering (GIWAXS) technique. In Figure 4i-l, all samples present diffraction rings, which indicate the samples' polycrystalline nature. Particularly, in comparison to the other perovskite thin films, for the sample with a FC-12 interlayer, the diffraction rings are sharp and speckled, indicating scattering from large crystallites. From the SEM and GIWAXS analysis, the perovskite grown on the FC-12 interlayer is highly crystalline with www.advmatinterfaces.de large grain size. Consequently, it can be concluded that defect states in FC-12-based samples are reduced, facilitating charge transport and separation, which enhances device efficiency and stability. Long term stability testing indicated that the FC-12 passivated device maintained ≈98% of its initial performance after exposure in dry air conditions for 1000 h, as opposed to 87% for samples without a passivation layer ( Figure S3, Supporting Information).
Thus, the influence of the FC-X length on the crystallization behavior of PbI 2 and perovskite grains has been investigated both theoretically and experimentally. In order to gain further insight into the charge transport mechanism, the charge transfer processes in the perovskite devices were studied in detail.
From the Schottky equation, the open circuit voltage of a solar cell could be expressed as follows [46][47][48] ln ln Where n is the ideality factor, k is the Boltzmann constant, T is the device temperature in Kelvin, q is electron charge, J sc is the short circuit current density and I is the light intensity, C is a constant and α is an empirical parameter which can be derived from the power function fitting of J SC against the illumination intensity. Thus, the magnitude of n could be calculated from the gradient of a suitable plot of light intensity versus V oc . Typically, the dominant recombination mechanism of the device under irradiation can be indicated from the n value, which varies from 1 to 2. For n = 1, the recombination is mainly related to band-band transitions and no trap-assisted recombination exists in the device. Alternatively, a value of n = 2 implies the presence of recombination via trap states within the band gap. Figure 5a shows the dependence of V OC on the light intensity in the devices with different interlayers. Linear fitting results indicate that the FC-12 device has the lowest slope (V OC versus the Napierian logarithm of the light-intensity) of 1.06 kT/q, corresponding to the lowest populated trap-state recombination, which is in agreement with the SEM and GIWAXS results when the perovskite is deposited on SnO 2 /FC-12. Conversely, the device with FC-2 treatment shows a much higher value of trap assisted recombination (1.99 kT/q), which may be caused by the poor crystal growth and surface macroscopic defects (pinholes). Devices using FC-6 and SnO 2 show a similar (1.75 kT/q) trap assisted recombination level with SnO 2 (1.72 kT/q) devices, which is higher than that in FC-12 devices. By fitting the LnJ sc and LnI values to a power function, α values of 0.98, 0.94, 0.94, and 0.99 were obtained for SnO 2 , SnO 2 /FC-2, SnO 2 /FC-6, and SnO 2 /FC-12-based devices, respectively (Figure 5b). Devices with SnO 2 /FC-12 as ETL exhibit the highest α value close to 1, implying that the bimolecular recombination is restrained, and free carriers were efficiently extracted to the electrodes before recombination could occur.
The electron transfer dynamics were further demonstrated by time-resolved PL (TRPL) spectra. TRPL is a valuable technique to describe the dynamics of various non-radiative and radiative decay pathways of photoexcited perovskite. Figure 5c www.advmatinterfaces.de shows the TRPL decays of perovskite thin films coated on various ETLs. A double-exponential function could be fit to the decays, and carrier lifetimes extracted. The lifetime and the corresponding proportions are listed in Table 2. In detail, the fitted lifetime consists of two components which imply different photophysical processes. The fast decay component (τ 1 ) arises from charge carrier quenching at the perovskite/ETL interface, while the slow decay component (τ 2 ) is induced by bimolecular recombination of free charge carriers originating from traps in the bulk. [38] The average decay lifetime (τ ave ) values of 34.0 and 19.6 ns were observed for perovskite samples on glass and SnO 2 ETL substrates, respectively, while the FC-12 sample shows the shortest τ ave value (11.1 ns). The fast component τ 1 made up the majority (83.7%) of the average decay lifetime, implying efficient electron transfer from perovskite to the ETL. The electron extraction capacity of different interlayers can be evaluated by the electron-transfer yield (Ф ext ), which can be estimated using the equation Ф ext = 1 − τ p /τ perov. , where τ p is the average lifetime for perovskite deposited on different substrates, and τ perov. is the average decay lifetime for initial perovskite thin film coated without any ETL. As expected, the SnO 2 /FC-12/perovskite thin film presented the highest electron transfer efficiency of 67.4%, compared to 42.3%, 35.0%, and 37.3% for SnO 2 , SnO 2 /FC-2, and SnO 2 /FC-6 ETLs, respectively. The immediate consequence of an optimized electron transfer yield is that an increased number of photon-generated electrons are extracted from the perovskite absorber into the external circuit, thus improving the J sc . Incident-photon-to-charge conversion efficiency (IPCE) plots were performed for confirmation. From Figure 5d it can be seen that, for the measured wavelength range, all the plots showed a similar trend, with IPCE showing a broad maximum from 380 to 750 nm, indicating that the influence of FC-X interlayer on the PSC device is mainly caused by facile electron transfer, rather than any enhanced absorbance at a particular wavelength. All the devices with FC-X interlayer displayed higher IPCE than the SnO 2 device with the FC-12 device presenting the highest IPCE, ≈90% across the broad maximum. Integrated current densities show 20.63, 21.48, 21.37, and 22.45 mA cm −2 for the SnO 2 , SnO 2 /FC-2, and SnO 2 /FC-6 PSCs, respectively. These results agree with the trend discussed above (TRPL fitted data) and measured values from the J-V curves.

Conclusions
In summary, we have designed and synthesized a series of amine-terminated fullerene derivatives with a gradedincreasing side carbon chain length (FC-2, FC-6, and FC-12) and applied them as the electron transfer interlayers in  perovskite solar cells. It was found that the alkyl-chain length plays a significant role in the device performance. A plausible interface passivation mechanism related to the amine-induced anchor crystallization process was developed and further demonstrated by contact angle and perovskite grain size variation. Compared to molecules with shorter side carbon chains, the amine-anchored crystallization process induced by the FC-12 molecule significantly improved device performance, attributed to the superior qualities of the perovskite when deposited on SnO 2 /FC-12 ETLs, such as larger grain size, enhanced crystallinity and reduced trap density. Planar perovskite solar cells with a high PCE of 21.2% were achieved by using the FC-12 long chain molecule. This work exemplifies the importance of crystallization engineering for perovskite solar cells and provides guidance for screening and design of other modifications in the future.
Synthesis of the FC-X Molecules: The side-chain functionalized fullerenes were made using standard literature methods [24] through the reaction of the respective diamines with PC 61 BM in high yields (95-97%). The experimental details are provided in the Supplementary Information.
Perovskite Solar Cells Fabrication: In a typical fabrication process, the ITO glass (sheet resistance 15 Ω cm −2 ) substrates were cleaned by sonicating in acetone, deionized water and IPA for 20 min, respectively. After drying the substrates by N 2 flow, the ITO substrates were placed in a UV-ozone chamber for 15 min before use. A SnO 2 nanoparticle solution diluted to 2.67% by deionized water was spin coated onto the ITO substrate at 4000 rpm for 30 s, followed by annealing in ambient air at 160 °C for 30 min. The above process was repeated once to obtain a compact SnO 2 thin layer. After another 15 min of UV-ozone treatment, the ITO/SnO 2 substrate was transferred into a N 2 filled glovebox. PbI 2 precursor was spin coated onto SnO 2 at 1500 rpm for 15 s and 5000 rpm for 20 s, and then annealed at 70 °C for 1 min. The PbI 2 solution was prepared by dissolving 1050 mg PbI 2 into 1.5 mL (DMF/DMSO = 10:1) mixed solvent with stirring overnight at 70 °C. Then, 0.3 mL of the mixture FAI: MABr: MACl (90 mg:11 mg:9 mg in 1 mL IPA) solution was spin coated onto the PbI 2 at 2000 rpm for 60 s, and a thermal annealing of 150 °C for 30 min in ambient air condition (30-40% humidity) was undertaken. For the interlayer coating, the FC-X molecule was dissolved into DCB with a concentration of 2 mg mL −1 , and stirred overnight at 60 °C, then 100 uL FC-X solution was dripped on the SnO 2 at 5000 rpm for 30 s. The solution was filtered using a 0.22 µm filter before use. The hole transporting layer was deposited on top of the perovskite layer at 3500 rpm for 30 s using 2,2′,7,7′-tetrakis(N,N-dip-methoxyphenylamine)-9,9′-spirobifluorene (Spiro-OMeTAD) solution, which consisted of 72 mg Spiro-OMeTAD in 1 mL chlorobenzene, with 18 µL bis(trifluoromethane) sulfonimide lithium salt (Li-TFSI) stock solution (520 mg Li-TFSI in 1 mL acetonitrile) and 28 µL 4-tertbutylpyridine (TBP) as dopant. Finally, a 100 nm Ag electrode was thermal evaporated as a counter electrode using a shadow mask with the electrode area of 0.1 cm. [2] A 0.045 cm 2 nonreflective metal mask was used to define the accurate active cell area during J-V curve measurement. The devices were kept in air with controlled humility (20%-35%), the J-V curves were measured once a week to monitor the stability of the device.
Characterization: The J-V plots were carried out in ambient air using a Keithley 2400 Source Meter under simulated one-sun AM 1.5G illumination (100 mW cm −2 ) with a solar simulator combined with digital exposure controller (Newport, US), and the light intensity was calibrated by means of a KG-5 Si diode. The devices were measured from 1.2 to 0 V, with a voltage step of 0.02 V.
The scanning electron microscopy (SEM) images were acquired by using field-emission scanning electron microscopy (FEI Teneo VolumeScope), with an electron beam accelerated at 20 kV. Absorption spectra were carried out via a UV-vis spectrometer (Cary 5000). Steady state PL was measured by an FLS980 spectrometer (Edinburgh, England). Time-resolved PL decays were measured by time-correlated single photon counting (TCSPC) microscopy, films were excited with a pulsed laser (PicoQuant PDL 800-B, λex = 465 nm). A long-pass filter separated excitation light from the emission and the latter was focused onto a fast response avalanche photodiode detector (ID Quantique, ID100). A timecorrelated single-photon counting card (PicoQuant, Time-Harp 200) was used to obtain the photoluminescence decay. Overall instrument response function FWHM was 200 picoseconds. The surface wettability of SnO 2 /FC-X was performed on a contact angle meter (OCA-20, Germany). The EQE was measured using a homemade EQE measurement system composed of a lamp (Newport 66 902) and monochromator (Newport 77 829). XPS data were acquired using a Kratos AXIS Supra X-ray photoelectron spectrometer equipped with a monochromated Al-Kα X-ray source (1486.7 eV) and a concentric hemispherical analyzer.
GIWAXS experiments were performed at the Australian Synchrotron on the SAXS/WAXS beamline under ambient conditions. A Pilatus 200K detector was used for 2D diffraction pattern collection. The energy of the incident beam was 12 keV at a range of incident angles from θ = 0.11-0.5°. The sample-to-detector range was 68.58 cm. Data from GIWAXS experiments was gathered from the SAXS/WAXS beamline of the Australian Synchrotron and analyzed using a customized version of NIKA 2D-based in IgorPro. Atomic force microscopy images were acquired using an Asylum Research Cypher scanning probe microscope operated in tapping mode.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.